Wireless sensor network (WSN) is a collection of large number of self-organized types of sensors which chain together to monitor and record physical or environmental conditions (i.e. used to measure temperature, sound, pressure) and passes gathered information to the central location. WSN build bridge between real world and virtual environment, which makes it more utilizable for many applications. Mainly WSN was used for military arena but now a days it is used in various area like industrial applications, consumer applications, health care applications and many more. Despite of having many advantages there are some issues also occurred in WSNs like hotspot problem, energy hole problem, routing, coverage problem, load balancing problem and so on. These issues effect on different factors of WSN named energy consumption, stability, quality, deployment time, lifetime of network, which degrade the performance of the WSN. To solve these issues various researchers develop different mechanisms. Among all of them, in this paper, we survey different kind of soft computing paradigms. Soft computing is a technique to use of improper solutions to solve the complicated problem in robust time. There are various types of soft computing techniques developed: swarm intelligence, fuzzy logic, neural network, reinforcement learning and evolutionary algorithm, which used to solve WSN problems so that performance of the network will be increased.